Radiomic Deep Brain Stimulation Prediction
^{with Quantitative Susceptibility Mapping (RadDBS-QSM)}

This repository hosts the following articles

Technical Feasibility of Quantitative Susceptibility Mapping Radiomics for Predicting Deep Brain Stimulation Outcomes in Parkinson’s Disease
published in Neurosurgery

Radiomic Prediction of Parkinson’s Disease Deep Brain Stimulation Surgery Outcomes using Quantitative Susceptibility Mapping and Label Noise Compensation
published in Brain Stimulation

and several conference papers.

Contents

Demonstration code can be found in main.ipynb
Radiomic features can be found in npy
Customizable extraction code is located in extract.py

Summary

A radiomic model based on presurgical quantitative susceptibility maps (QSM) is used to predict patient outcomes to deep brain stimulation (DBS) surgery for the treatment of Parkinson's disease.

Model overview.

This project presents a framework to:

• Extract radiomic features for input into a regression model to predict post-surgical motor improvement.
  
• Incorporate clinical variables such as age, sex, etc.
• Provide a novel label noise compensation technique improving outcome prediction.
  

Installation

Clone the repository with

git clone https://github.com/agr78/RadDBS-QSM.git

Navigate to the repository

cd RadDBS-QSM

Run the setup script

source ./install.sh

Wait...then open the Jupyter notebook in the RadDBS-QSMenv environment

jupyter notebook ./src/jupyter/main.ipynb

Notes

• This tool was developed for use with QSM, but can be used with other contrasts.
• If the QSM has not been reconstructed, this repository provides code to obtain the whole brain susceptibility.
• If manual region-of-interest masks are not available, this repository provides bash scripts to create a sample atlas and register individual cases.

Publications

If this code is used, please cite the following:

Neurosurgery Article: A. G. Roberts et al., "Technical Feasibility of Quantitative Susceptibility Mapping Radiomics for Predicting Deep Brain Stimulation Outcomes in Parkinson’s Disease, 2025, DOI: 10.1227/neu.0000000000003721

BibTex

@article{Roberts_RadDBS-QSM_2025,
  title    = "Technical feasibility of quantitative susceptibility mapping
              radiomics for predicting deep brain stimulation outcomes in
              Parkinson disease",
  author   = "Roberts, Alexandra G and Zhang, Jinwei and Tozlu, Ceren and
              Romano, Dominick and Akkus, Sema and Kim, Heejong and Sabuncu,
              Mert R and Spincemaille, Pascal and Li, Jianqi and Wang, Yi and
              Wu, Xi and Kopell, Brian H",
  journal  = "Neurosurgery",
  month    =  sep,
  year     =  2025,
  keywords = "Deep brain stimulation; Machine learning; Parkinson disease;
              Quantitative susceptibility mapping; Radiomics; Regression",
  language = "en"
}

Contact

Please direct questions to Alexandra G. Roberts at agr78@cornell.edu.